Summary

Leading AI researchers increasingly argue that artificial intelligence has become an essential technology and should no longer be left to major tech corporations. Instead of relying on a few dominant providers, they call for the development of public AI infrastructure that creates space for smaller models and local innovations. Countries like Switzerland with the Apertus project and Finland with LUMI demonstrate how state investments can democratize access to AI technology.

People

Topics

  • Democratization of AI technology
  • Public AI infrastructure
  • Open-source AI models
  • Data security and data protection
  • Regional AI specialization

Detailed Summary

Dominance of Tech Giants and Counter-Movement

Currently, a few major corporations like OpenAI and Google dominate the AI market. The necessary financial and technological resources to train large language models are unattainable for small businesses. However, this is leading to a growing counter-movement: scientists argue that AI, as a transformative technology, is better placed in public hands – at least partially.

Smaller Models as a Solution

Yejin Choi emphasizes that smaller, specialized AI models are entirely sufficient and often even more cost-effective. These could be developed at lower cost while simultaneously addressing critical data protection concerns – particularly in sensitive areas like healthcare. Smaller open-source models enable organizations to control AI solutions locally without having to share sensitive data with large tech corporations.

Swiss Initiative: Apertus

ETH Zurich developed Apertus, an open AI model that has been freely available since summer 2025 and openly discloses its training data. Professor Andreas Krause emphasizes that public institutions should not develop commercial products themselves, but can create an innovative breeding ground – a foundation on which Swiss startups and companies can build.

Finland's LUMI Project

Finland goes even further. The LUMI project creates state AI infrastructure with a high-performance supercomputer – more powerful than ETH's Alps supercomputer in Lugano. This enables small startups to gain access to computing resources they could never afford themselves.

European Strategy: Specialization Instead of Superintelligence

Petri Myllymäki from LUMI argues against the goal of universal superintelligence. Instead, Europe should develop thousands of specialized AI models for concrete problems – for drug development, power grids, health data analysis. LUMI is one of 19 similar projects in the European Union designed to put Europe on the right path.


Key Takeaways

  • AI as a public good: AI is too important to remain entirely in private hands.
  • Small models are practical: Specialized, smaller AI models are more cost-effective and more data protection-friendly.
  • Public infrastructure is central: State investments in AI infrastructure democratize access to technology.
  • Regional specialization: Europe should not attempt to build a generic "superintelligence," but rather specialized solutions for local challenges.
  • Switzerland and Finland as pioneers: Projects like Apertus and LUMI show that small to medium-sized countries can take innovative approaches.

Stakeholders & Those Affected

Who is affected?Who benefits?Who loses?
Society, small/medium enterprises, public institutionsStartups, local ecosystems, data protection, competitionTech giants (monopolistic market position)

Opportunities & Risks

OpportunitiesRisks
Reduction of tech monopoliesHigher public investments required
Better data protection through local controlFragmentation into hundreds of incompatible models
Innovation in regional ecosystemsTechnological dependence on infrastructure operators
Specialized, problem-oriented AI solutionsGeopolitical tensions (USA vs. Europe)
Transparency and traceabilitySecurity risks with open-source models

Relevance for Action

Decision-makers should:

  1. Finance public AI infrastructure – similar to national research centers.
  2. Support open-source initiatives – through research funding and access to computing resources.
  3. Define data standards – to ensure security and compatibility.
  4. Promote regional specialization – instead of global competition for "super-AI".
  5. Create regulatory frameworks – that don't overburden small, specialized models.

Quality Assurance & Fact-Checking

  • [x] Central statements and figures verified
  • [x] Quotes extracted directly from source text
  • [x] Projects (Apertus, LUMI, Alps) verified
  • [x] No unconfirmed data included
  • [ ] ⚠️ Publication date WEF Davos 2026 based on article metadata: 27.01.2026

Additional Research


Bibliography

Primary Source:
AI as a Public Good – Is AI Too Important to Leave to Tech Corporations? – Sandro Della Torre, SRF (26.01.2026)
https://www.srf.ch/wissen/kuenstliche-intelligenz/ki-als-oeffentliches-gut-ist-ki-zu-wichtig-um-sie-den-tech-konzernen-zu-ueberlassen

Supplementary Sources:

  1. ETH Zurich – Apertus: Open AI Model (2025)
  2. LUMI Supercomputing Project – European AI Infrastructure (EuroHPC, Finland)
  3. WEF Davos 2026 – Discussion program on AI and public goods

Verification Status: ✓ Facts checked on 26.01.2026


Footer (Transparency Notice)


This text was created with the support of Claude.
Editorial responsibility: clarus.news | Fact-checking: 26.01.2026